Daniel's post has got me thinking..... I've got a ...
# modules
n
Daniel's post has got me thinking..... I've got a few bioinformatics scripts that I've built into modules using a standard dockerfile that I push up in a github action to an internal registry..... I've elected to use a dockerfile pattern similar to the following: https://github.com/nf-core/modules/blob/master/modules/nf-core/bcl2fastq/Dockerfile but I'm wondering if I should adapt my build process to look more like the following: https://github.com/nf-core/modules/blob/master/modules/nf-core/ensemblvep/Dockerfile as it would simplify a migration to Wave if my organization wanted, is that accurate?
d
For me, using the
environment.yml
-- or another Conda specification -- we have in nf-core modules make more sense. No need to create a Dockerfile, just run
wave
with
environment.yml
(not forgetting
--freeze
) seems simpler.
n
Gotcha ok thank you! I agree the Wave pattern is very desirable and having the forward compatibility is a plus. I think perhaps internally we can evaluate this on a per tool basis. Since we aren't using Wave right now the need isn't strong but I do like using conda for R tooling because build times in R are much better when using Conda.
d
For R, and other programming language-based tools, Wave -- actually Seqera containers (the
--freeze
param) -- is perfect as I see it, as all I need to do is create a single file specifying all libraries which can then be used in Nextflow for Conda, Docker and Singularity/Apptainer. And I can tailor make environment files for any purpose; no need to make a mulled Biocontainer containing any library I might ever use in future. However, I'm very much a rookie in this field so I can't say I understand everything in detail.
👍 1